This document reports on the annotation and evaluation experiment of 
the DIT++ tagset that has been carried out in January-March 2006 at 
Tilburg University.  It presents statistical results for each dimension 
in the tagset based on the annotation data and elaborates on 
peculiarities found in the annotations by looking to both cooccurence 
matrices and examples. The findings are enriched by feedback from some of 
the annotators, making this report suitable for getting an impression about 
the 'usability' of the tagset in human annotation, and at the same time 
providing a starting point for minor revisions on the tagset. This report 
concludes by stating that in general and on average, the agreement scores on 
the dimensions based on the current scheme seem sufficient for using the 
tagset to describe functions in dialogue. Furthermore, it is concluded that 
for few dimensions there is substantial room for improvement in definition 
and explanation. It is expected that this will result in higher agreement 
than reported here.